ContentOS is a multi-tenant platform built for the real needs of an agency running a dozen industry blogs at once. It combines keyword research (GSC/GA4/Merchant), an AI pipeline with 5 models and auto-publishing to 6 Magento instances. The flow is configurable step by step — which model does the research, which writes the intro, which edits. Brand context is injected into every prompt automatically, so an article about pellets for the Austrian market never talks about Polish subsidies.
For every step in the flow you pick a specific provider, model and prompt. Perplexity handles research because it works in real time, Gemini drafts because of the long context, Claude edits because the prose reads naturally, Higgsfield makes the featured image. This is not an "auto-writer" — it is a pipeline you shape around your own product.
Every number comes from production code — verifiable on a call under NDA. As of 2026-05-09 (local agent report).
Tenant switcher at the top (8 to choose from). Three key metrics: Articles, Flows, Pipeline Runs. Below that, the latest articles with their status (draft / review / published) and the latest pipeline runs (completed / failed / running) — you know immediately what needs attention.
The full article table for a single tenant. Filters: active / excluded from SEO, status, type (pillar / standalone), domain, publication, gaps, language. Every article has its cluster, keywords (tags), version history, publication date and last update. Export to Magento in one click.
The content flow editor: Keyword Research → TOP 10 SERP analysis → Topic research → Content Brief → Pillar article. Every node has its own provider (Perplexity, Claude, Gemini) and a specific model — you change it in the right-hand sidebar. A custom prompt per step, with brand context auto-injected.
The engine scans the existing content plus GSC queries plus the competition and surfaces 50 active alerts: content gaps, critical omissions (GDPR, costs, availability), topics in content decay, an AI report on trending topics. Every alert is clickable → an article draft lands in the pipeline.
An annual strategy with 99 strategic topics — each with an SEO score, traffic potential, phase (Content / Technical) and category. You click "Approve" → the topic goes into the pipeline. It takes "what shall we write today" out of the client briefing.
Quality analysis across the whole content set: SEO score, Performance, Quick Wins, Content Gaps. Data imported from GSC, automatic categorisation per article (Critical / Warnings / Good). Quick Wins = articles one position away from the top 3 — a concrete list of what to fix this week.
ContentOS imports products straight from Magento (8 products in the screenshot — Black Friday, Senior Add-on, Extra Minutes, Carer, Senio.Bis and so on). The product audit detects critical omissions: no meta description, no description, no images, names that are too short. Every issue is clickable → a fix prompt goes to the AI.
The full link table: source, link URL, anchor, type (article / product / category), status. The AI suggests new links based on per-field similarity (title + intro + body scored separately). You can insert links in bulk or run a "Link Audit" — it checks for broken links and duplicate tracking.
Once an article is published, ContentOS automatically generates post variants for Facebook, LinkedIn and Instagram — with emoji, hashtags, a CTA and an image. You can approve, edit or reject each one. Auto-publishing to a Facebook Page runs through the Facebook Page MCP API.
Each article generates a featured image plus in-article images automatically (Gemini text-to-image + Higgsfield image-to-image). All assets in one gallery, per tenant. You can upload your own, generate new ones from a prompt, clone and edit.
ContentOS pulls the available models dynamically from each provider's /api/models. A new model ships — it appears in the dropdown. Switching provider mid-flow is one click, no deployment.
Editing, natural prose, long context. Default choice for Polish and German copy.
Drafting, long articles, text-to-image. Cheapest at 100k+ tokens.
SEO meta tags, reasoning, structured outputs. Fallback for edge cases.
Realtime web + X (Twitter) signals. Trending topics, news angles.
Real-time research with citations. Default choice for fact-finding and competitor analysis.
Most content marketing tools stop at "generate a draft". ContentOS goes further — it publishes, measures, optimises and checks for cannibalisation. Because we built it for ourselves, as an agency running 8 industry blogs.
You choose provider + model + prompt for every step of the flow. Research on Perplexity, draft on Gemini, editing on Claude — or any other combination. Save a preset per content category.
One click publishes to the right store view with the correct slug, the featured image in pub/media/ and an HTML wrapper. Supports 6 Magento instances plus a webhook to Next.js (opiekunek).
A crawl of the client's site and blog, an AI-generated brand profile, context injected into every prompt. The article "knows" where it will be published — brand voice, target market, linking policy.
Three layers: URL similarity, embedding similarity, query overlap from Google Search Console. It shows where two articles are fighting over the same phrase — before publication, not six months later.
Suggests links based on per-field similarity (title + intro + body scored separately). It updates older articles by adding links to new ones. Built-in anchor-text rules for brand and commercial terms.
Gemini text-to-image for featured and in-article images. Higgsfield image-to-image to place the client's products in a new setting — boiler room, delivery, warehouse. No stock photos of people in suits.
Per-tenant A/B testing of SEO meta with CTR tracking in GSC. After 30 days the system picks the winning variant automatically. It takes "I think this sounds better" out of SEO optimisation.
A publishing calendar showing which phrases are already in the pipeline. A SERP rank tracker per country with alerts when positions drop. A bulk product editor for shops (Magefan + Magento PRD).
8 tenants in production. Each has its own brand context, its own integrations and its own API keys for the AI providers. Postgres Row Level Security — zero crosstalk, zero chance of a leak between clients.
Postgres with Row Level Security on every table. Every query is filtered automatically by tenant_id. The per-tenant site context is injected at runtime into every AI prompt.
9 integrations in production. Some are simple (REST), some are a nightmare (Magefan returns string-in-string JSON, reports errors as HTTP 200, and produces slug conflicts across stores). Every one of them has been repaired — the commit history proves it.
"8 domains, 5 AI models, one publishing API. Content goes from research into Magento with no spreadsheets and no copy-paste."
The standard StudioApps stack plus Inngest for durable workflows instead of cron and a Redis queue. Inngest gives you retries per step, observability and replay — and there is no Redis to maintain.
ContentOS is an internal tool — we do not sell it as SaaS. There is no public onboarding, no billing, no plans. Active users: around 3 (the StudioApps team). The 8 tenants are clients under NDA plus our own brands.
ContentOS is our internal tool today — but we are opening it selectively to partner agencies. A no-obligation call: 30 minutes, a screenshare from our production account, a real pipeline running on live domains.